Forecasting Work Conditions for Road Construction Activities: An Application of Alternative Probability Models

Abstract

Certain activities of highway construction are particularly sensitive to such weather conditions as soil moisture, precipitation, and daily temperature. Regression analysis is used to obtain three alternative probability models designed to translate observed weather conditions into probabilities for carrying out construction activities. The models were developed using generalized least squares, normit analysis, and logit analysis. The generalized least squares method was the most convenient computationally, but it had severe interpretative disadvantages. The results obtained by logit analysis gave the desired probabilistic interpretation most readily and had the best predictive ability. Comparison of sample observation and predicted work probabilities for common excavation during wet and dry months indicated that the logit analysis model could accurately translate weather conditions into probabilities that work would take place. Models for paving and asphalt work and for bridge and drainage structure are also estimated using logit analysis. These estimates indicate a strong sensitivity of the latter category of work to precipitation conditions. Such models may aid contract letting agencies is planning payment schedules, penalty clauses, and completion dates for new roads; construction firms may find such models valuable in planning effective use of men and equipment.